getdoc0b9a. 157KB Jun 04 2011 12:04:04 AM

J
ni
o
r
Elect

o

u

a
rn l

c

o

f
P

r

ob
abil
ity

Vol. 6 (2001) Paper No. 24, pages 1–14.
Journal URL
http://www.math.washington.edu/~ejpecp/
Paper URL
http://www.math.washington.edu/~ejpecp/EjpVol6/paper24.abs.html

LINEAR STOCHASTIC PARABOLIC EQUATIONS,
DEGENERATING ON THE BOUNDARY OF A DOMAIN
S. V. Lototsky
University of Southern California
Department of Mathematics
1042 Downey Way., DRB 142
Los Angeles, CA 90089-1113
lototsky@math.usc.edu
Abstract A class of linear degenerate second-order parabolic equations is considered in

arbitrary domains. It is shown that these equations are solvable using special weighted
Sobolev spaces in essentially the same way as the non-degenerate equations in R d are
solved using the usual Sobolev spaces. The main advantages of this Sobolev-space approach are less restrictive conditions on the coefficients of the equation and near-optimal
space-time regularity of the solution. Unlike previous works on degenerate equations, the
results cover both classical and distribution solutions and allow the domain to be bounded
or unbounded without any smoothness assumptions about the boundary. An application
to nonlinear filtering of diffusion processes is discussed.
Keywords Lp estimates, Weighted spaces, Nonlinear filtering.
AMS subject classification 60H15, 35R60.
This work was partially supported by the NSF grant DMS-9972016.
Submitted to EJP on November 30, 2000. Accepted October 17, 2001.

1

Introduction

Sobolev spaces Hpγ are very convenient to study parabolic equations in all of

Rd :

∂u
= aij Di Dj u + bi Di u + cu + f,
∂t

(1.1)

where summation over the repeated indices is assumed from 1 to d. Roughly speaking,
γ+1−2/p
if the initial condition is in Hp
and the right hand side is in Hpγ−1 , p ≥ 2, then
γ+1
u ∈ Hp as long as the coefficients are bounded and sufficiently smooth, and the matrix
(aij ) is uniformly positive definite. It was shown in [4] that a similar result holds for the
Ito stochastic parabolic equations
du = (aij Di Dj u + bi Di u + cu + f )dt + (σ ik Di u + ν k u + g k )dwk

(1.2)

as long as the stochastic right hand side g k is in Hpγ and the matrix (aij − (1/2)σ ik σ jk ) is
uniformly positive definite.

The objective of this paper is to show that, if the Sobolev spaces Hpγ are replaced with
weighted spaces, then an analogous result holds for the Ito stochastic parabolic equations
with quadratic degeneracy of the characteristic form. Let ρ = ρ(x) be a smooth function
so that ρ(x) ∼ dist(x, ∂G) near the boundary. Consider a linear stochastic parabolic
equation
du = (ρ2 aij Di Dj u + ρbi Di u + cu + f )dt + (ρσ ik Di u + ν k u + g k )dwk .

(1.3)

Equations with operators of the type ρα ∆, α > 0, have been studied by many authors
in deterministic setting [9, 13, 14], and operators with quadratic degeneracy of the characteristic form (α = 2) always required separate treatment. Therefore, in the stochastic
setting, it is also natural to consider these operators separately.
To study equation (1.3), the Sobolev spaces Hpγ are replaced with certain weighted Sobolev
γ
spaces. These weighted space Hp,θ
(G) were first introduced in [5] to study stochastic
parabolic equations on the half-line, and further investigated in subsequent papers by
γ
both authors. In the notation Hp,θ
(G) of the space, the indices γ, p have the same meaning

γ
as in Hp , and the index θ determines the boundary behavior of the functions from the
space: the larger the value of θ, the faster the functions and their derivatives can blow up
γ
near the boundary of G. The advantage of solving equation (1.3) in the space Hp,θ
(G) is
that existence and uniqueness results can be obtained for a wide class of linear and quasilinear equations. Unlike many related works in the deterministic setting, these results, for
different values of γ and θ, cover both classical and distribution solutions and, for γ > 0,
go beyond abstract solvability. Indeed, for γ > 0, embedding theorems show that the
solution is a continuous function of x and t and, for sufficiently large p, has almost equal
number of classical and generalized derivatives.
γ
Since the spaces Hp,θ
(G) have been used in the analysis of the Dirichlet boundary value
problem for nondegenerate parabolic equations [6, 5, 7], let us recall the main result.

2

Consider the Dirichlet problem for equation (1.2) in a sufficiently regular domain. Suppose
that the coefficients are sufficiently smooth, the matrix (aij − (1/2)σ ik σ jk ) is uniformly

γ+1−2/p
γ−1
γ
positive definite inside the domain, and f ∈ Hp,θ
(G), g k ∈ Hp,θ
(G), u|t=0 ∈ Hp,θ
(G),
γ+1
p ≥ 2. Then, for certain values of θ, the solution will be in Hp,θ (G). Note that θ of the
solution space is different from the corresponding values for the initial condition and the
right hand side, and if θ is too large or too small, then the corresponding solvability result
does not hold.
The results are quite different, and completely analogous to the whole space, if we consider degenerate parabolic equations. Namely, for equation (1.3), the solution will be in
γ+1
Hp,θ
(G) as long as the coefficients are sufficiently smooth, the matrix (aij − (1/2)σ ik σ jk )
γ−1
γ
is uniformly positive definite inside the domain, and f ∈ Hp,θ
(G), g k ∈ Hp,θ

(G),
γ+1−2/p

u|t=0 ∈ Hp,θ
(G), p ≥ 2. Now, the result holds for all real θ, and the function ρ
can be chosen so that no restrictions are necessary about the domain G. The domain can
be bounded or unbounded, without any smoothness of the boundary, and this generality
makes the results new even in deterministic setting.
Recall [10, Chapter 5] that the stochastic characteristic for equation (1.3) is the diffusion
process x = xt defined by
dxt = −ρ(xt )B(t, xt )dt + ρ(xt )r(t, xt )dW (t)

(1.4)

with an appropriate choice of r, B, and W . Assuming that the functions ρ, B, r are globally
Lipschitz continuous and bounded, the unique solvability of (1.4) implies that, if x0 is in
G, then xt will never reach the boundary of G. This is why it is natural to expect that
the solvability results for (1.3) will not involve any conditions about the boundary of G.
γ
The construction and analysis of the spaces Hp,θ

(G) for general domains are in [8]. Section
2 presents a summary of results from [8] and the construction of the necessary stochastic parabolic spaces. The main result of the paper, the statement about solvability of a
second-order degenerate semi-linear stochastic parabolic equation, is in Section 3. In Section 4, an application is given to the problem of nonlinear filtering of diffusion processes,
when the unobserved process evolves in a bounded region.

In this paper, notation D m is used for a generic partial derivative of order m with respect
to the spatial variable x = (x1 , . . . , xd ); Di = ∂/∂xi . Summation over repeated indices is
assumed.

2

Definition of the spaces

Let G ⊂ R d be a domain (open connected set) with non-empty boundary ∂G. Denote by
ρG (x), x ∈ G, the distance from x to ∂G. For n ∈ Z define the subsets Gn of G by
Gn = {x ∈ G : 2−n−1 < ρG (x) < 2−n+1 }.
3

(2.1)

Let {ζn , n ∈ Z} be a collection of non-negative functions with the following properties:
X
ζn (x) ≥ δ > 0.
(2.2)
ζn ∈ C0∞ (Gn ), |D m ζn (x)| ≤ N(m)2mn ,
n∈Z
The function ζn (x) can be constructed by mollifying the characteristic (indicator) function
of Gn . If Gn is an empty set, then the corresponding ζn is identical zero.
Recall [13, Section 2.3.3] that the space Hpγ is defined for γ ∈ R and p ≥ 1 as the
completion of C0∞ (R d ) with respect to the norm k · kHpγ = kΛγ · kLp (Rd ) , where Λγ f =
((1 + |ξ|2 )γ/2 fˆ )ˇ, andˆ,ˇare the Fourier transform and its inverse. Similarly, Hpγ (l2 ) is the
set of sequences g = {g k , k ≥ 1} for which
kgkHpγ (l2 ) := k kΛγ gkl2 kLp (Rd ) < ∞,
where kgkl2 =

P

k≥1 |g

k 2 1/2
|

(2.3)

.

Definition 2.1 Let G be a domain in R d , θ and γ, real numbers, and p ∈ [1, +∞). Take
a collection {ζn , n ∈ Z} as above. Then
)
(
X
γ
Hp,θ
2nθ kζ−n (2n ·)u(2n ·)kpHpγ < ∞ ,
(2.4)
(G) := u ∈ D ′ (G) : kukpH γ (G) :=
p,θ
n∈Z
where D ′ (G) is the set of distribution on C0∞ (G);

)
(
X
γ
Hp,θ
2nθ kζ−n (2n ·)u(2n·)kpHpγ (l2 ) < ∞ .
(G; l2 ) := u ∈ D ′ (G; l2 ) : kukpH γ (G;l2 ) :=
p,θ
n∈Z
(2.5)
γ
A detailed analysis of the spaces Hp,θ
(G) is given in [8]. In particular, it is shown that
γ
Hp,θ
(G) does not depend on the particular choice of the system {ζn } and, for p > 1, is a
reflexive Banach space.

Remark 2.2 If G is a bounded domain, then summation in (2.4) is carried out over
n ≤ n0 for some integer n0 depending on the domain. In particular, if G is bounded, then
γ
γ
Hp,θ
(G) ⊂ Hp,θ
(G) for θ1 > θ2 .
2
1
Definition 2.3 Let ρ = ρ(x), x ∈ R d , be a function so that
1. ρ(x) = 0, x ∈
/ G;
m+1
2. ρ is infinitely differentiable in G, and |ρm
ρ(x)| ≤ N(m) for all x ∈ G and
G (x)D
for every m = 0, 1, . . ..

3. N1 ρG (x) ≤ ρ(x) ≤ N2 ρG (x) for some N1 , N2 > 0 and all x ∈ G.
4

Functions introduced in the above definition do exist; for example,
X
ρ(x) =
2−n ζn (x).

Z

(2.6)

n∈

The conditions in the above definition imply that ρ is uniformly Lipschitz continuous in

R d ; in particular, this is true for the function defined by (2.6). On the other hand, if G is
a bounded domain and the boundary ∂G is of class C |γ|+2 , γ ∈ R , then, by Lemma 14.16
in [1], it is possible to choose the function ρ so that ρ ∈ C |γ|+2 (R d ).
For ν ≥ 0, define the space Aν (G) as follows:
1. if ν = 0, then Aν (G) = L∞ (G);
2. if ν = m = 1, 2, . . . , then
m−1 m−1
m−1
Aν (G) = {a : a, ρG Da, . . . , ρG
D
a ∈ L∞ (G), ρm
a ∈ C 0,1 (G)},
GD

kakAν (G) =

m−1
X

m−1
kρkG D k akL∞ (G) + kρm
akC 0,1 (G) ;
GD

(2.7)
(2.8)

k=0

3. if ν = m + δ, where m = 0, 1, 2, . . . , δ ∈ (0, 1), then
m
δ
m
ν
Aν (G) = {a : a, ρG Da, . . . , ρm
G D a ∈ L∞ (G), ρG D a ∈ C (G)},

kakAν (G) =

m
X

kρkG D m akL∞ (G) + kρνG D m akC δ (G) .

(2.9)
(2.10)

k=0

Theorem 2.4 1. Assume that γ − d/p = m + ν for some m = 0, 1, . . . and ν ∈ (0, 1). If
γ
u ∈ Hp,θ
(G), then
m
X
k=0

γ
sup |ρk+θ/p (x)D k u(x)| + [ρm+ν+θ/p D m u]C ν (G) ≤ N(d, γ, p, θ)kukHp,θ
(G) .

(2.11)

x∈G

Recall that [f ]C ν (G) = supx,y∈G |x − y|−ν |f (x) − f (y)|.

2. Given γ ∈ R define γ ′ so that γ ′ = 0 for integer γ and γ ′ is any number from (0, 1)
γ
as long as |γ| + γ ′ is not an integer for non-integer γ. Then, for every u ∈ Hp,θ
(G) and
′
|γ|+γ
a∈A
(G),
γ
γ
ka ukHp,θ
(2.12)
(G) ≤ N(γ, p, θ, d)kakA|γ|+γ ′ (G) kukHp,θ
(G) .
γ
These and other properties of the spaces Hp,θ
(G) and Aγ (G) can be found in [8].

Definition 2.5 Fix (Ω, F , {Ft}, P ), a stochastic basis with F and F0 containing all P | τ ]] = {(ω, t) ∈ Ω × R + : 0 < t ≤ τ (ω)}; P, the
null subsets of Ω; τ , a stopping time, (0,
σ-algebra of predictable sets; {wk , k ≥ 1}, independent standard Wiener processes. The
Ito stochastic integral will be used.
5

The following Banach spaces were introduced in [4] to study stochastic parabolic equations
on R d :
| τ ]]; P; Hpγ ), H γ
| τ ]]; P; Hpγ (l2 ));
1. H γp (τ ) = Lp ( (0,
p (τ ; l2 ) = Lp ( (0,
γ+1−2/p

2. Fpγ (τ ) = H pγ−1 (τ ) × H γp (τ ; l2 ), Upγ = Lp (Ω; F0 ; Hp

);

3. Hpγ (τ ): the collection of processes from H γ+1
(τ ) that can be written, in the sense of
p
distributions, as
Z t
Z t
u(t) = u0 +
f (s)ds +
g k (s)dwk (s)
(2.13)
for some u0 ∈

Upγ

and (f, g) ∈

0
γ
Fp (τ );

0

kukpHγp (τ ) = kD 2 ukpH γ−1 (τ ) + k(f, g)kpFpγ (τ ) + Eku0 kp γ+1−2/p .

(2.14)

Hp

p

For a positive real number T > 0, a stopping time τ ≤ T , a real number δ ∈ (0, 1], and a
(Banach space) X -valued process u, we will use the following notation:
kukpC δ ([0,τ ],X)

= sup ku(t ∧
0≤t≤T

τ )kpX

+

sup
0≤s 0, there exists δε > 0 so that
for all (ω, t) ∈ (0,

|ρG (x)aij (t, x) − ρG (y)aij (t, y)| + kρG (x)σ i· (t, x) − ρG (y)σ i· (t, y)kl2 ≤ ε

(3.4)

| τ ]] and x, y ∈ G with |x − y| < δε . See Theorem 2.4(2) for the definition
for all (ω, t) ∈ (0,
of γ ′ .

Assumption 3.3 (Regularity of the free terms.)
γ
(f (·, ·, 0, 0), g(·, ·, 0)) ∈ Fp,θ
(τ, G),

(3.5)

and for every ε > 0 there exists µε > 0 so that
γ
k(f (·, ·, u, Du) − f (·, ·, v, Dv), g(·, ·, u) − g(·, ·, v))kFp,θ
(τ,G)

≤ εku − vkHγp,θ (τ,G) + µε ku − vkH γp,θ1 (τ,G) , γ1 < γ + 1, (3.6)
γ
for all u, v ∈ Hp,θ
(τ, G).

7

γ
Definition 3.1 A process u ∈ Hp,θ
(τ, G) is a solution of (3.1) if and only if the equality

u(t, x) = u0 (x) +

Z

0

t

(ρ2 (x)aij (s, x)Di Dj u(s, x) + f (s, x, u, Du))ds
Z t
+
(ρ(x)σ ik (s, x)Di u(s, x) + g k (s, x, u))dwk (s) (3.7)
0

γ
holds in Hp,θ
(τ, G).
γ
Theorem 3.2 If p ≥ 2, τ ≤ T , and u0 ∈ Up,θ
(G), then, under Assumptions 3.1–3.3,
there is a unique solution u of equation (3.1) and


p
p
p
p
kukHγ (τ,G) ≤ N · k(f (·, ·, 0, 0)kH γ−1 (τ,G) + kg(·, ·, 0)kH γ (τ,G;l2 ) + Eku0 k γ+1−2/p
p,θ

p,θ

p,θ

Hp,θ

(G)

(3.8)

with the constant N depending only on γ, κ1 , κ2 , p, T, θ, and the functions ρ, δε , µε .
Proof. The arguments are very similar to the proof of Theorem 3.2 in [7]. A more
detailed description of the method can be found in [3, Sections 6.4,6.5].
To simplify the presentation, assume that τ = T and introduce the following notations.
Define the operators
Au(t, x) = ρ(x)aij (t, x)Di Dj u(t, x), Bk u(t, x) = ρ(x)σ ik (t, x)Di u(t, x)

(3.9)

γ
γ
γ
and write (|A, B|)u = (f, g, u0) for some (f, g) ∈ Fp,θ
(T, G), u0 ∈ Up,θ
(G) if u ∈ Hp,θ
(T, G)
Rt
Rt k
k
and u = u0 + 0 (Au + f )ds + 0 (B u + g )dwk (s).

Let {ζn , n ∈ Z} be the collection of functions used in Definition 2.1 and let {ηn , n ∈ Z}
be a collection of functions so that ηn ∈ C0∞ (Gn ), |D m ηn (x)| ≤ N(m)2mn , ηn (x) = 1
on the support of ζn . Using Theorem 5.1 in [4], for (ϕ, ψ) ∈ Fpγ (T ) and v0 ∈ Upγ , define
Sn (ϕ, ψ, v0 ) ∈ Hpγ (T ) so that v = Sn (ϕ, ψ, v0 ) if and only if v ∈ Hpγ (T ), v|t=0 = v0 , and
dv = (η−n ρ2 aij Di Dj v + 22n (1 − η−n )∆v + ϕ)dt + (η−n ρσ ik Di v + ψ k )dwk (t).

(3.10)

Note that


kv(·, 2n ·)kHγp (T ) ≤ N · k(ϕ(·, 2n·), ψ(·, 2n ·))kFpγ (T ) + kv0 (2n ·)kUpγ

(3.11)

with N independent of n. Indeed, for a function f = f (x) write fn (x) = f (2n x). Then
vn (t, x) = v(t, 2n x) satisfies
dvn = (22n η−n,n ρ2n aij
n Di Dj vn + (1 − η−n,n )∆vn + ϕn )dt
+ (2n η−n,n ρn σnik Di vn + ψnk )dwk (t), (3.12)
where η−n,n (x) = η−n (2n x). Since ρ ∼ 2n on the support of η−n , inequality (3.11) follows
from Theorem 5.1 in [4].
8

γ
Note also that, for every u ∈ Hp,θ
(T, G),

Sn ((|A, B|)(ζ−n u)) = ζ−n u.

(3.13)

Assume
that f and g depend only on t and x. Also, with no loss of generality, assume
P first
2
that n ζn (x) = 1 for all x ∈ G. If (|A, B|)u = (f, g, u0), then it follows from (3.13) that
X
u=
ζ−n Sn (ζ−n f − An u, ζ−ng − Bn u, ζ−n u0 ),
(3.14)
n

where An u = A(uζ−n) − ζ−n Au, Bnk u = Bk (ζ−n u) − ζ−n Bk u.

γ
γ
Conversely, for every (f, g) ∈ Fp,θ
(T, G) and every u0 ∈ Up,θ
(G), equation (3.14) has
γ
a unique solution u ∈ Hp,θ (T, G) so that u satisfies (3.8) and (3.1) (recall that so far
we assume that f, g do not depend on u). Indeed,
by inequalities (3.11) and (2.21),
P
a sufficiently high power of the operator u 7→
n ζ−n Sn (An u, Bn u, 0) is a contraction
γ
in Hp,θ (T, G), which implies the existence of a unique solution of (3.14). This solution
satisfies


Z T
p
p
p
p
kukHγ (T,G) ≤ N · k(f, g)kF γ (T,G) + ku0kU γ (G) +
kukHγ (t,G) dt ,
p,θ

p,θ

p,θ

0

p,θ

γ
and (3.8) follows by the Gronwall inequality. Since u ∈ Hp,θ
(T, G), we have (|A, B|)u =
γ
(f0 , g0, u0 ) for some (f0 , g0 ) ∈ Fp,θ (T, G), and it follows from (3.14) that f¯ = f − f0 , g¯ =
P
g − g0 satisfy n ζ−n Sn (ζ−n f¯, ζ−n g¯, 0) = 0. Applying the operator (|A, B|) to the last
equality and using (3.13), we conclude that
X
X
f¯ =
An Sn (ζ−n f¯, ζ−n g¯, 0), g¯ =
Bn Sn (ζ−n f¯, ζ−n g¯, 0).
n

n

Once again, using inequalities (3.11) and (2.21), we conclude that a sufficiently high power
of the operator
!
X
X
(f, g) 7→
An Sn (ζ−n f, ζ−n g, 0),
An Sn (ζ−n f, ζ−n g, 0)
n

n

is a contraction, which means that (f¯, g¯) = (0, 0) and u is a solution of (3.1) with f, g
independent of u.

γ
γ
Now, for every (f, g) ∈ Fp,θ
(T, G) and every u0 ∈ Up,θ
(G) we can define u = R(f, g, u0 ) ∈
γ
Hp,θ (T, G) so that (|A, B|)u = (f, g, u0). This means that every solution of (3.1) with
general (f, g) satisfies u = R(f (u, Du), g(u), u0). To conclude the proof of the theorem,
we note that Assumption 3.3 implies that a sufficiently high power of the operator u 7→
γ

R(f (u, Du), g(u), u0) is a contraction in Hp,θ
(T, G). Theorem 3.2 is proved.

Corollary 3.3 Assume that τ = T , the initial condition u0 is compactly supported in
γ+1−2/p
G and belongs to Hp
, and Assumption 3.3 holds for all θ ∈ R . If p > 2 and
γ + 1 − (d + 2)/p > m for some positive integer m, then the solution u(t, x) of (3.1) has
the following properties:
9

1. for almost all ω ∈ Ω, u is a continuous function of (t, x);
2. for almost all ω ∈ Ω and all t ∈ [0, T ], u is m times continuously differentiable in
¯ as a function of x;
G
3. for almost all ω ∈ Ω and all t ∈ [0, T ], u(t, x) and its spatial partial derivatives of
order less than or equal to m vanish on the boundary of G.
Proof. By assumption, u0 ∈ Hp,θ (G) for every θ ∈ R , because compactness of support
of u0 means that the corresponding sum in (2.4) contains finitely many non-zero terms.
γ
Consequently, by Theorem 3.2, u ∈ Hp,θ
(G, T ) for all θ ∈ R . It remains to use (2.22) with
β sufficiently close to 1/p, and then apply Theorem 2.4(1).

2−2/p

Remark 3.4 The linear equation
du(t, x) = (ρ2 (x)aij (t, x)Di Dj u(t, x) + ρ(x)bi (t, x)Di u(t, x) + c(t, x)u(t, x) + f (t, x))dt
+ (ρ(x)σ ik (t, x)Di u(t, x) + ν k (t, x)u(t, x) + g k (t, x))dwk (t) (3.15)
γ
satisfies the hypotheses of the theorem if (f, g) ∈ Fp,θ
(τ, G), the functions bi , c, ν k are
P ⊗ B(G) measurable, and

kbi (t, ·)kAnb (G) + kc(t, ·)kAnc (G) + kν(t, ·)kAnν (G;l2 ) ≤ κ2 .

(3.16)

As for the values of nb , nc , nν , we can always take nb = |γ| + γ ′ , nν = nc = |γ + 1| + γ ′ ,
but these conditions can be relaxed. For example (cf. [4, Remark 5.6]), if γ ≥ 1, we can
take nb = nc = γ − 1 + γ ′ , nν = γ + γ ′ .

4

Application to nonlinear filtering of diffusion processes

The classical problem of nonlinear filtering is considered for a pair of diffusion processes
(Xt , Yt ) defined in R d1 by equations
dXt = b(t, Xt , Yt )dt + r(t, Xt , Yt)dWt
dYt = B(t, Xt , Yt )dt + R(t, Yt )dWt

(4.1)

with some initial conditions X0 , Y0. It is assumed that Wt is a d1 -dimensional Wiener
process on a complete probability space (Ω, F , P ), Xt is d-dimensional state process,
d1 > d, and Yt is (d1 − d)-dimensional observation process. The coefficients are known
functions of corresponding dimension and are smooth enough for a unique strong solution
to exist. The filtering problem consists in computing the conditional density of Xt given
the observations up to time t. It is known [10, Chapter 6] that under some natural
regularity assumptions, this conditional density satisfies a nonlinear stochastic parabolic
10

equation, also know as Kushner’s equation. Alternatively, the density can be computed by
normalizing the solution of the Zakai equation, which is a linear equation. Both analytical
theory and numerical methods for the Kushner and Zakai equations have been studied by
many authors, and these studies made (4.1) the standard model in filtering theory.
Nonetheless, for most applications, (4.1) is only an approximation. There are two main
reasons for that. First, the actual process Xt usually evolves in a bounded region, for
example, because of mechanical restrictions, and therefore equations (4.1) are a suitable
model only when the process Xt is away from the boundary. Second, even if Xt evolves
in the whole space, the corresponding filtering equations, when solved numerically, are
considered in a bounded domain, which effectively restricts the range of Xt . Therefore,
it seems natural to start with a filtering model in which the state process evolves in a
bounded region. The model presented below is just one possible way to address the issue
of replacing the whole of R d with a bounded domain. The model can be easily analysed
using the theory developed in the previous sections of the paper, but it is certainly not
the most general filtering model in a bounded domain.
Let G be a bounded domain and ρ, a scalar function as in Definition 2.3. Consider the
following modification of the classical filtering model:
dxt = ρ(xt )b(t, xt , yt )dt + ρ(xt )r(t, xt , yt)dWt
dyt = B(t, xt , yt )dt + R(t, yt )dWt

(4.2)

with some initial conditions x0 , y0 . Since ρ is Lipschitz continuous in R n , by uniqueness
of the solution of (4.2), the process xt can never cross the boundary of G. Note that G is
any bounded domain. It is shown below (Lemma 4.1) that, if the domain G is sufficiently
large, x0 ∈ G, and the function ρ is chosen in a special way, then (xt , yt ) are close to
(Xt , Yt ).
For a matrix M denote by M ∗ its transpose. We make the following assumptions. For
the discussion of these assumptions see Section 8 in [4].
Assumption 4.1 The functions b, B, r, R are bounded and Borel measurable in (t, x, y)
and uniformly Lipschitz continuous in (x, y). The function r = r(t, x, y) is continuously
differentiable with respect to x and the derivatives are continuous in y and uniformly
Lipschitz continuous in x.
Assumption 4.2 The matrix RR∗ is invertible and V = (RR∗ )−1/2 is a bounded function
of (t, y).
Assumption 4.3 There exists δ > 0 so that, for all (x, y) ∈
ξ ∈ Rd ,
(r(1 − R∗ V 2 R)r ∗ ξ, ξ) ≥ δ|ξ|2.

R d × R d1 −d , t > 0, and all
(4.3)

Assumption 4.4 The initial condition (x0 , y0 ) is independent of Wt , the conditional
distribution of x0 given y0 has a density Π0 , and Π0 is compactly supported in G.
11

Lemma 4.1 Let {GK , K > 0} be a collection of domains with smooth boundaries so
that BK := {x : |x| < K} ⊂ GK and dist(∂GK , BK ) > δ for some fixed δ > 0. For
each GK , let ρK be a function as in Definition 2.3 so that ρK (x) = 1, x ∈ BK , and
N1 ρGK (x) ≤ ρK (x) ≤ N2 ρGK (x), x ∈ GK , N1 , N2 are independent of K, x. Define the
K
processes x = xK , y = y K according to (4.2) with ρ = ρK , xK
0 = X0 I(X0 ∈ BK ), y0 = Y0 ,
p
p
and assume that E(|X0 | + |Y0| ) < ∞ for some p > 0. Then, as K → ∞, sup0≤t≤T |Xt −
K
xK
t | + sup0≤t≤T |Yt − yt | converges to zero with probability one and in every Lq (Ω), q < p.
K p
Proof. First of all, notice that E sup0≤t≤T (|Xt | + |xK
t | + |Yt | + |yt |) ≤ C with C
independent of K. Indeed, for example,
Z t
K
sup |yt | ≤ |Y0 | + CT + sup |
R(s, ysK )dWs |
0≤t≤T

0≤t≤T

0

and it remains to use Burkholder-Davis-Gundy inequality [2, Theorem IV.4.1].
K
Define random variable ηK = sup0≤t≤T |Xt − xK
t | + sup0≤t≤T |Yt − yt |. Since ρ(x) = 1 for
|x| ≤ K, equations (4.1) and (4.2) imply
\
{ω : lim ηK > 0} ⊆
{ω : sup |Xt | > N},
K→∞

N ≥1

0≤t≤T

and then, by the Chebuchev inequality,
P(

\

E sup0≤t≤T |Xt |p
= 0.
N →∞
Np

{ sup |Xt | > N}) = lim P ({ sup |Xt | > N}) ≤ lim

N ≥1

0≤t≤T

N →∞

0≤t≤T

q
After that, since the family {ηK
, K > 0} is uniformly integrable for q < p [11, Lemma
II.6.3],
q
q
EηK
= EηK
I( sup |Xt | > K) → 0, K → ∞.
0≤t≤T



Introduce the following notations:
a(t, x) = (1/2)rr ∗(t, x, yt ) ∈ R d×d , σ(t, x) = rR∗ V (t, x, yt ) ∈ R d×(d1 −d) ,
h(t, x) = V B(t, x, yt ) ∈ R d1 −d , ht = h(t, xt ),
L[v] = Di Dj (ρ2 aij v) − Di (ρbi v), Λk [v] = hk v − Di (ρσ ik v).

(4.4)

The filtering problem for (4.2) is to find conditional distribution of xt given the observations {ys , 0 < s ≤ t}. Since the function ρ is zero outside of G, equations (4.2) can be
considered in the whole R d . Then, by Theorem 8.1 in [4], the conditional expectation of
f (xt ) givenR {ys , 0 < s ≤ t}, for every bounded measurable function f = f (x), can be
written as Rd f (x)Π(t, x)dx. By the same theorem, the function Π = Π(t, x) belongs to
Hp1 (T ), for very T > 0, and is the solution of a nonlinear equation
X
¯ k Π)(V k RdWt + (hk − h
¯ k )dt)
dΠ = L[Π]dt +
(Λk [Π] − h
(4.5)
t
t
t
k≤d1 −d

12

¯t =
with
initial condition Π0 , where V k is the kth row of the matrix V and h
R
Rd h(t, x)Π(t, x)dx. It is often convenient to consider the un-normalized filtering density u = u(t, x) given by a linear equation
X
du = L[u]dt +
Λk [u](V k RdWt + hkt dt)
(4.6)
k≤d1 −d

with initial condition u|t=0 = Π0 , so that
u(t, x)
Π(t, x) = R
.
u(t,
x)dx
d
R

(4.7)

Theorem 4.2 Suppose that G is a bounded domain and Assumptions 4.1 –4.4 hold. If
2−2/p
1
Π0 belongs to Lp (Ω; Hp
), for some p ≥ 2, then both u and Π belong to Hp,θ
(T, G) for
all T > 0 and all θ ∈ R .
Proof. By Assumption 4.4, Π0 ∈ Hp,θ (G) for every θ ∈ R , because compactness of
support of Π0 means that the corresponding sum in (2.4) contains finitely many non-zero
1
terms. Consequently, Theorem 3.2 and Remark 3.4 imply that u ∈ Hp,θ
(T, G).
2−2/p

1
To show that Π ∈ Hp,θ
(T, G), we first use Theorem 8.1 in [10], according to which the
solution Π of (4.5), considered in the whole space, belongs to Hp1 (T ). Then by Theorems
4.2.2 and 4.3.2 in [12] we find

kζ−n (2n ·)Π(·, 2n ·)kpH1p (T ) ≤ NkΠ(·, 2n ·)kpH1p (T ) ≤ N2β|n| kΠkpH1p (T ) ,

(4.8)

where β and N are positive and independent of n. Remark 2.2 then implies that Π ∈
1
¯ t as a known process
Hp,θ
(T, G) for sufficiently large θ1 . On the other hand, by treating h
1
and applying Theorem 3.2 with γ = 1 and θ < θ1 , we conclude that (4.5) has a unique
1
solution belonging to Hp,θ
(T, G). By uniqueness, this solution is Π.

If the conditions of Theorem 4.2 hold for sufficiently large p, then, by Corollary 3.3,
for all t ∈ [0, T ] and almost all ω ∈ Ω, the conditional density Π(t, x) is continuously
¯ and both Π(t, x) and its first-order partial derivatives with respect to
differentiable in G
x vanish on the boundary of G.

13

References
[1] D. Gilbarg and N. S. Trudinger. Elliptic Partial Differential Equations of Second Order.
Reprint of the 1998 edition, Springer, Berlin, 2001.
[2] N. V. Krylov. Introduction to the Theory of Diffusion Processes. American Mathematical
Society, Providence, RI, 1995.
[3] N. V. Krylov. Lectures on Elliptic and Parabolic Equations in H¨
older Spaces. American
Mathematical Society, Graduate Studies in Mathematics, v. 12, Providence, RI, 1996.
[4] N. V. Krylov. An analytic approach to SPDEs. In B. L. Rozovskii and R. Carmona,
editors, Stochastic Partial Differential Equations. Six Perspectives, Mathematical Surveys
and Monographs, pages 185–242. AMS, 1999.
[5] N. V. Krylov and S. V. Lototsky. A Sobolev Space Theory of SPDEs with Constant
Coefficients on a Half Line. SIAM J. Math. Anal., 30(2):298–325, 1999.
[6] N. V. Krylov and S. V. Lototsky. A Sobolev Space Theory of SPDEs with Constant
Coefficients in a Half Space. SIAM J. Math. Anal., 31(1):19–33, 2000.
[7] S. V. Lototsky. Dirichlet Problem for Stochastic Parabolic Equations in Smooth Domains.
Stoch. Stoch. Reports, 68(1-2):145–175, 1999.
[8] S. V. Lototsky. Sobolev Spaces with Weights in Domains and Boundary Value Problems for
Degenerate Elliptic Equations. Methods and Applications of Analysis, 7(1):195–204, 2000.
[9] O. A. Oleinik and E. V. Radkeviˇc. Second Order Equations with Nonnegative Characteristic
Form. AMS and Plenum Press, Rhode Island, New York, 1973.
[10] B. L. Rozovskii. Stochastic Evolution Systems. Kluwer Academic Publishers, 1990.
[11] A. N. Shiriaev. Probability. Springer, New York, 1984.
[12] H. Triebel. Theory of Function Spaces II. Birkhauser, Basel, 1992.
[13] H. Triebel. Interpolation Theory, Function Spaces, Differential Operators. Johann Amrosius
Barth, Heidelberg, 1995.
[14] M. I. Viˇsik and V. V. Gruˇsin. Boundary Value Problems for Elliptic Equations Which are
Degenerate on the Boundary of the Domain. Math. USSR Sb., 9:423–454, 1969.

14